NDTnet - December 1997, Vol.2 No.12
Low and Multi-Frequency Eddy Current Techniques
Assure the Integrity of RPV Clad
in Nuclear Power Plants
Dr. Reiner Becker, Dr. Ludwig von Bernus
Dr. Michael Disqué, Prof. Dr. Michael Kröning,
Michael H. Dalichow
Some German nuclear power plants are approaching the end of their designed life
spans. Considering the possibility of radiation damage to the RPV material in close
proximity to the core, reliable nondestructive testing techniques for the detection,
classification, and sizing of surface and sub-surface cracks, as well as sub-clad cracking
with extensions into the base material, are of immense interest. It has been shown that
weld microstructure is particularly sensitive to neutron degradation (primarily due to
high copper content), and thus very stringent NDE safety assessments are imperative.
In addition to inspecting clad integrity, clad thickness measurements also become
essential. This paper discusses the development of optimized, high dynamic-range,
eddy current techniques utilizing a specialized yoke-type absolute coil transmitter and
differential coil receiver. The eddy current working frequencies are below 500 Hz, and
the combination of eddy current data with data acquired using a multi-frequency
approach allows for signal-to-noise ratio enhancements and filtering of disturbing
signals. Results are presented from the qualification process, which documented the
reliability of the system as confirmed with specific test and calibration blocks, and
during inservice inspections. And finally, because of the presented eddy current
techniques in combination with optimized ultrasonic testing techniques, two older
German reactors were able to achieve a high level of safety and confidence, and were
once again put on-line.
2. NDE Inspection Task
According to German nuclear power safety guidelines, the internal austenitic steel clad
of reactor pressure vessels is assumed to be a protective layer to prevent corrosion of
the carbon steel pressure vessel. Therefore, material properties and thickness of the
clad have never been considered as contributors to strength and/or toughness in design
and fracture mechanical regards. Inservice NDT inspection requirements only call for
the inspection of the bond between clad and base material using straight beam
longitudinal wave techniques. The volume of the clad is not inspected; however,
ultrasonic testing is used for detecting sub-clad flaws. Typically, this is done using dual
element transmit/receive search units to inspect the sub-clad material with refracted
longitudinal waves (usually 70°) at a focal point distance of 30 mm (1.2 inch).
The development and validation of new NDT techniques occurred with two goals in
mind: 1) To supplement standard techniques with diverse information to assist in
interpretation of potential nonconformance; and 2) To provide information to evaluate
aspects of component lifetime extension.
Currently in Germany, lifetime extension issues are not of actual concern. Recently,
however, the continued operation and subsequent licensing of Germany's oldest
pressurized water reactor (400 MW, two loops) was under discussion. The primary
concern was an assumed increase of neutron degradation of circumferential welds (in
the belt-line region) and field welds with increased copper content filler material. Due
to this condition and in accordance with the specific license directives, the clad areas of
and adjacent to the welds have to be inspected during each outage using a multi-frequency
eddy current technique. This technique employs frequencies above 20 kHz
and is limited to detection of surface-connected flaws. An optimized mixing algorithm
 was applied to suppress disturbing signals caused by sensor lift-off at the rough and
uneven cladding surface and by local -ferrite variations.
Considering the clad contribution to overall strength and ductility when applying
fracture mechanics, it was demonstrated that imbrittlement can be excluded. To secure
this approach, the NDT techniques selected to document the integrity of the cladding,
including detection and characterization of flaws in the volume of the clad and at the
interface with the base material, had to be validated and qualified.
3. Research and Validation Program
The proposed research emphasized expansion of contemporary eddy current
technology into detection and characterization of flaws in the clad volume, or beyond
the penetration depth of traditional eddy currents. The clad to base material interface
and the sub-clad base material areas were to be inspected using enhanced ultrasonic
techniques. Due to the acoustic properties of manual or automated deposit clad, the
limited ability of ultrasound to detect small flaws (reflectors) is a well-known problem;
classification of reflectors as planar flaws is generally restricted in these particular areas.
To solve this well-known problem, the utilization of low-frequency  and multi-frequency
eddy current techniques was proposed, because these techniques offer an
improvement in detection ability for small planar flaws. Contrary to the limitations of
ultrasound, the proposed eddy current technique has many benefits. With eddy
current, the classification, characterization, and sizing of reflector indications in the clad
volume and sub-clad area are generally possible. Critical flaws, such as planar flaws,
originating at the clad to base material interface and extending into the base material
(sub-clad cracking) can be distinguished from those located in the clad volume that
were caused during the manufacturing process.
3.1 Inspection System and Techniques
The new system is based on low-frequency eddy current testing  that is
complemented with a high-frequency component, providing a large frequency
spectrum that is more practical in eddy current testing. This allows for an all-inclusive
and detailed NDT assessment of the clad , .
The system employs the following features:
Figure 1: Multi-Frequency
Eddy Current System
Figure 2: Hardware Components
Wide signal dynamic range (> 85 dB) and strong long-range stability, for
detection of even the smallest measurement effects within a high-interference
- Effective computing ability using integrated digital signal processors (DSP).
- Powerful signal processing algorithm.
Figure 1 shows the inspection system. The analog eddy current electronic is contained
in the interface box to the right of the PC, which is used for data acquisition,
presentation, and documentation.
The analog boards and the DSP-board are the main components contained in the
interface box, as depicted in Figure 2 below. The DSP-board can combine up to four
analog boards to accommodate different operation modes, such as a time-parallel mode
with four frequencies, time-multiplexing mode, or multi-frequency mode. The transfer
of inspection data to the DSP-board and to the PC is performed via a serial interface.
The interrelation of the hardware and software is presented in Figure 3. The basic
concept is to control most of the system functions numerically by the digital signal
processing software in the DSP to minimize hardware components and expense.
The signal processing firmware is based on a multi-regression analysis. The available
algorithms can be interpreted as a numerical filter for different inspection tasks.
Figure 4: Calibration Variables
- Least-Square Fit
- Suppression of Disturbing (non-relevant) Signals
- Flaw Detection, Classification, Characterization & Sizing
- Collecting all Elements Required for "Calibration Data-Matrix"
- Calculation of Filter Coefficients of the Different Functions
- On-Line Signal Mix, using Distinct Filters to Suppress Noise
and Non-Relevant Input Data
The objective of the signal processing is to suppress disturbing signals, which allows for
the quantitative evaluation of target functions (i.e., the characterization of flaw type and
size). The calibration of the inspection system requires well-defined calibration and/or
test blocks, which represent the complete set of relevant component parameters,
including all known component condition variations found during the actual
Different frequency ranges combined with either absolute or differential sensors allow
inspection for a variety of flaw types or clad conditions, as presented below in Figure 5.
Figure 5: Inspection Techniques
|High-Frequency - Absolute Sensor (50/280/500 kHz)
- Surface Connected Flaws
- Surface Profiling
- -Ferrite Content and Distribution
|Low-Frequency - Absolute Sensor (500/2800/5000 Hz)
- Clad Thickness Measurements
|Low-Frequency - Differential Sensor (500/2800/5000 Hz)
- Subsurface Volumetric Flaws
- Subsurface Planar Flaws
- Sub-Clad Flaws
3.2 Results Obtained on Calibration Blocks and Test Specimen
When inspecting RPV clad, a number of deviant parameters have to be considered that
make processing of specific data quite difficult. On the other hand, comprehensive
information of the clad condition can be individually filtered for each of the inspection
tasks listed in Figure 5, using an appropriate firmware/software package. The
following Figures below present actual inspection results obtained using three
Figure 6: Surface Profiling, HF-Absolute Sensor
Figure 7: -Ferrite Content, HF-Absolute Sensor
Figure 8: Surface Connected Planar Flaws, HF-Absolute Sensor
Figure 9: Clad Thickness, LF-Absolute Sensor
Figure 10: Subsurface Volumetric Flaws, LF-Differential Sensor
Figure 11: Planar Flaws in the Clad Volume, LF-Differential Sensor
Figure 12: Sub-clad Planar Flaws, LF-Differential Sensor
The display of post-evaluation results using a numerical filter optimized for the
measurement of lift-off between the sensor and the clad surface is depicted in Figure 6.
The color-coded C-scan image on the left side closely represents the surface profile of
the scanned surface. The grooves of the strip-welded clad are clearly displayed along
with a localized repair (ground-out). The numerical amount (in millimeters) of the
sensor lift-off is indicated in the lower A-scan presentation at the current index-line
The spatial distribution of the -ferrite content of the scan area is displayed in Figure 7
on the following page. The numerical filter for the -ferrite prediction is applied to the
acquired raw data. Both A-scans indicate the percentage of the -ferrite content at the
cross-line cursor position.
To detect and analyze the scan area for surface-connected planar flaws, a filter is
applied that suppresses non-relevant signals caused by the lift-off effect and/or signals
due to -ferrite influences. The C-scan, presented in Figure 8, shows the indications of
surface-connected EDM notches contained in a calibration block. These notches---
oriented parallel, transverse, and oblique to the weld---are easily detected and
displayed in the C-scan presentation. The depth of each individual notch is indicated in
the amplitude display (A-scan display) as a function of amplitude height. Additional
filters can be applied to select the display of planar reflectors of specified depth. To
obtain the required quality of the above presented results (lift-off filter, -ferrite filter,
and surface-connected planar flaw filter), high-frequency techniques, operating with 50
kHz, 280 kHz, and 600 kHz are used for inspection.
For investigation of the clad volume, low-frequency techniques using 510 Hz, 2.8 kHz,
and 5 kHz are employed. Figure 9 presents clad thickness results furnished with the
numerical filter for clad thickness testing. Local thickness variations are clearly
indicated in the C-scan presentation, where clad thickness ranges from 6 mm up to 12
mm, whereas the nominal clad thickness should be 8 mm.
For the detection and characterization of volumetric flaws buried in the clad volume,
specific differential coil sensors operating in the low frequency range are applied.
Figure 10 displays evaluation results using a numerical filter for the detection of
volumetric flaws in the clad volume. The C-scan presentation reveals three side-drilled
holes. On top, the cursor location on the A-scan presentation shows the indication
amplitudes representing remaining ligaments of 2, 3, 4, and 6 millimeters as a function
of amplitude height. Again, additional filters can be applied to select for display of only
reflectors of a specified depth. Planar flaw indications in the clad volume and flaws in
the sub-clad volume are not detected with this filter type; these require application of a
filter described in the following paragraph.
Figure 11 presents results of the inspection for planar flaws in the clad volume,
applying yet another specific numerical filter. All unwanted signals, such as those
described previously, are suppressed so as to display planar flaws embedded in the
clad only. The cursor location on the A-scan presentation at the top shows the
amplitudes of the indications representing remaining ligaments of 4 and 6 mm as a
function of amplitude height. The filter applied for this flaw type can be expanded to
select the display of planar reflectors of specified depth also.
Finally, a numerical filter to detect and characterize planar flaws in the sub-clad volume
(under-clad cracking) is applied, as presented in Figure 12. The indication is detected
at a remaining ligament of 8 millimeters, representing the clad to base material interface
at nominal clad thickness. Due to the local increase of electrical conductivity at the clad
to base material interface, this inspection technique benefits from the enhancement of
the eddy current density, producing a high signal-to-noise ratio for the detection of
planar flaws extending to the clad to base material interface.
3.3 Inservice Inspection Experience
The system, as described above, was applied in 1995 during Spring and Fall outages in
two aging German nuclear power plants. Much practical knowledge was gained from
this experience, including that the available set of calibration and test blocks
represented only a limited, insufficient amount of possible component variables,
specifically for the real-world RPV clad. Local variations of the clad condition could be
detected, especially at repair locations; however, it was learned that relatively large
values for sensor lift-off and sensor-tilt have to be considered, particularly for older
plants with excessive grooves between the weld strips.
Our experience showed that calibration of the numerical filter functions using plant-specific
calibration blocks (that should closely represent the plant-specific component
variables and clad deposit process) produced relevant indications along with
indications (false calls) which could not be replicated on the given calibration block(s).
Prompted by these "false call" indications, we improved our software to extend the
calibration from the standard calibration block(s) reflectors to additional calibration
reflectors obtained from the actual RPV clad. The resulting evaluation software can be
applied off-line after the data acquisition process.
The approach for the development of extended calibration procedures can be described
The filter coefficients obtained from the standard calibration block(s) are applied to
the actual component input data. Obviously, only indications corresponding to the
flaw type and component condition of the calibration block(s) used during the filter
calibration can be analyzed. All other signals have to be analyzed without
correlation to known calibration block reflectors. Based on this situation, data
analysis is unreliable if sufficient correlation of relevant conditions and properties
of the calibration block(s) and inspected component is unknown.
For ISI without relevant baseline data, according to our experience, a sufficient
correlation of calibration block vs. component cannot be demonstrated. The set of
calibration data can be expanded (successive data expansion) using actual data
from the component, to disposition non-relevant indications using the following
The existing database is expanded with a few additional data points. These
additional data points (e.g., 50 additional positions) are randomly inserted in the
original inspection area of approximately 100,000 data points, thus decreasing
the general noise level. However, the ability of the algorithm for flaw detection,
classification, and sizing has to be controlled and verified on the standard
calibration block, and new recording thresholds have to be established. If,
accidentally, one of the additional data points is caused by a relevant flaw, the
statistical impact to the overall calibration is negligible.
The existing database is expanded with one or more of the "false calls",
regardless of relevancy, and tagged as non-relevant noise. This new filter data
is compared to the calibration data reflector of the standard calibration block
and the acquired component data. If the "false call" indication was caused by
non-relevant noise or disturbance, the signal-to-noise ratio of the component
data will be largely improved while the calibration data remain nearly
unchanged. If the "false call" signal from the component data was produced by
a relevant flaw response, then the new filter coefficient produces significant and
notable variations to the calibrated reference reflectors in the original calibration
data. A flaw, comparable in type and size with the calibrated data, can then be
This strategy, which significantly reduced the total number of reportable indications
without decreasing inspection quality, has been sanctioned by the German authorities
and it was successfully applied in the two older German nuclear power plants.
The documented results demonstrate the potential of the newly designed eddy current
inspection system to assess the integrity of reactor pressure vessel clad. Using the data
interpretation capabilities provided by the system's hard- and software, comprehensive
information in addition to more conventional NDE information is obtained. By virtue
of the inspection features, to characterize surface-connected flaws, volumetric and
planar flaws buried in the clad volume, and sub-clad flaws connected to the clad to base
material interface, combined with information pertaining to the surface topography,
cladding thickness, and -ferrite content and distribution, this new technology, when used in combination with ultrasonic and visual inspections, will provide significant
weight when discussing lifetime extension strategies for nuclear power plants. The
application of this technology (applied to the entire clad surface from the ID of the
reactor pressure vessels) in two aging German nuclear power plants provided major
contributions to the authority's decision-making process, which ultimately allowed the
continued operation of one plant and the restarting of the other. The assumed
imbrittlement of the reactor pressure vessels were safely excluded as they were shown
not to present any risks to the plant or to public safety and health.
- Libby, H.L.; Introduction to Electromagnetic Nondestructive Test Methods, 1st Ed.,
Wiley Interscience, New York, USA, 1971, chapter 7 and 8.
- Becker R., Disqué M.; Low frequency eddy current inspection of the pressure vessel
clad, new approach and future works in modeling, Proceedings of the EPRI Pipe
Inspection Workshop, Session V, Charlotte, USA, 1993.
- Becker, R., Both N., Kröning, M.; A New DSP-Controlled Eddy Current System In
Modular And Compact Design, Some Applications Of Reactor Component Flaw
Inspection, Proceedings of the EPRI Vessel & Internals Inspection Conference,
Session IIb, Charlotte, USA, 1994.
- Becker, R., Dobmann, G., Kröning, M., Reiter, H., Schneider, E.; Integration of NDT
into life-time management, Proceedings of the International Conference on Pipes
and Pressure Vessels, Feb 1 - 2, 1996, Singapore, Malaysia. To be published in The
International Journal of Pressure Vessels and Piping.
|Top to this page|
For more information see backup issue: NDTnet 12/97: NDT in Power Generation.
© Copyright 1. Dec 1997 Rolf Diederichs,
/DB:Article /AU:Bernus_L_v /AU:Becker_R /AU:Disque_M /AU:Kroening_M /AU:et_al /IN:FHG /CN:DE /CT:ET /CT:instrument /CT:energy /ED:1997-12